Claude is now doing a quarter of the work developing Anthropic’s next AI


Full story

Anthropic is using its Claude artificial intelligence model to build the next generation of AI. The company says Claude is doing more and more of the work developing the models that will eventually succeed it — and that shift has happened fast.

QR code for SAN app download

Download the Straight Arrow app today to get the stories that matter free from manipulation, bias or agenda.™

Point phone camera here

Back in February, Claude was leading basically none of Anthropic’s model research and development. Six months later, it was leading 26%. Anthropic says that means Claude can take a high-level prompt and do most of the work on a task from start to finish.

Humans still supervise it, so Claude isn’t building its replacement entirely on its own. But Anthropic says about 90% of its research and development now involves some level of collaboration with Claude.

Anthropic calls for slower AI development

Anthropic CEO Dario Amodei has been one of the people calling for a slowdown in AI development over safety concerns. The company now says models that help accelerate its own development could eventually become harder for humans to understand or control.

Dario Amodei, CEO and co-founder of Anthropic (AP Photo/Markus Schreiber, File)

“We should do everything possible to minimize the gap between what frontier labs know and what the public knows,” the company said in a blog post. “This means better measuring the development of AI, reporting on it publicly, and giving society an opportunity to decide how to use this information.”

Anthropic is also calling on other AI companies to share those metrics, as well. The company says doing so could help “understand how close the world is to reaching recursive self-improvement” — when a model can fully autonomously build its successor.

Hackers use Claude to access OpenAI

A separate Wall Street Journal report says independent security researchers used Claude while breaking into an OpenAI employee’s ChatGPT account, giving them access to private company software and the ability to suggest changes to it.

The researchers were participating in OpenAI’s bug bounty program, which pays security researchers to find vulnerabilities and report them to the company.

Round out your reading

Tags: , , , , , , , ,

Straight Arrow
Fear No Fact.™

Don't just take our word for it.


Center-rated reporting

According to media bias experts at AllSides

AllSides Center-rated reporting May 2026

Transparent and credible

Awarded a perfect reliability rating from NewsGuard

100/100

Welcome back to trustworthy journalism.

Find out more

Why this story matters

AI systems are now actively involved in building their own successors, and a security breach exposed private OpenAI software to outside researchers using a competing AI tool.

AI building itself faster

Anthropic says Claude went from leading none of its R&D in February to leading 26% now, with 90% of research involving some Claude collaboration.

Control concerns acknowledged

Anthropic's own blog post warns that AI models accelerating their own development could eventually make those systems harder for humans to understand or control.

OpenAI systems accessed

According to a Wall Street Journal report, outside researchers used Claude to access an OpenAI employee's ChatGPT account and read private company software.

Straight Arrow
Fear No Fact.™

Don't just take our word for it.


Center-rated reporting

According to media bias experts at AllSides

AllSides Center-rated reporting May 2026

Transparent and credible

Awarded a perfect reliability rating from NewsGuard

100/100

Welcome back to trustworthy journalism.

Find out more